Biochemistry 1993,32, 7593-7604
7593
Two-Dimensional Rotational-Echo Double Resonance of V a l 1 - r 1-13C]Gly2-[ 1 5 N ] A l a 3 - G r a m i c i d i n A in Multilamellar DimyristoylphosphatidylcholineDispersionst Andrew W. Hing' and Jacob Schaefer Department of Chemistry, Washington University, St. Louis, Missouri 63130 Received October 9, 1992; Revised Manuscript Received April 21, 1993
I3C-l carbon and Ala3 ISN-amidenitrogen was used to investigate the conformation and dynamics of the Glyz-Ala3 13CJ5N peptide bond in Vall-[ 1-l3C]Gly2[1SN]Ala3-gramicidin A incorporated into multilamellar dispersions of dimyristoylphosphatidylcholine. Measurement of the I3C-lSN dipolar coupling constant D of the labeled gramicidin in a powder and the effective dipolar coupling constant De in a multilamellar dispersion was accomplished by two-dimensional rotational-echo double-resonance (2DREDOR) NMR, a magic-angle spinning experiment designed to measure weak dipolar coupling constants. The magnitudes of D and De were measured by the mirrorsymmetric form of 2D REDOR, and the signs of D and De were determined relative to the sign of the isotropic indirect spin-spin coupling constant J by the mirror-asymmetric form of 2D REDOR. From knowledge of the magnitudes of D and De, four possible values were calculated for the angle between the Glyz-Ala3 l3C-I5N peptide bond and the gramicidin helical axis. Additional knowledge of the signs of D and De permitted the set of possible values for the peptide bond angle to be reduced to a single angle and its supplement (64O, 116O). This information about the Glyz-Alas 13C-lSN peptide bond angle eliminates the double-stranded, helical dimers and the left-handed, single-stranded, @ helical dimer but supports the right-handed, single-stranded, @.3 helical dimer as the structural model for gramicidin in multilamellar dispersions. ABSTRACT: The dipolar coupling between the Glyz
Val]-gramicidin A is a linear pentadecapeptide that has the following primary structure (Sarges & Witkop, 1965):
of these studies provide strong evidence that gramicidin forms a single-stranded, p3helical dimer (Urry, 1971; Urry et al., 1971; Arseniev et al., 1986; Venkatachalam & Urry, 1983) HCO-L-Val, -Gly,-~-Ala~-~-Leu,-~-Ala,-~-Val,-~-Val,in oriented bilayers and that the helices are right-handed D-Val,-L-Trp,-D-Leu,,-L-Trp, ,-D-LeU,,-L-Trp,,-D-LeUl~(Nicholson & Cross, 1989). L-Trpls-NHCHzCHzOH Fewer solid-state NMR experiments have been performed on gramicidin in multilamellardispersions. In these systems, Because gramicidin forms ion channels in model membranes the isotropic distribution of molecular orientations generates (Hladky & Haydon, 1970, 1972), gramicidin has been so-called powder pattern spectra. Powder patterns represtudied as a potential model of ion channels in cell membranes. senting motionally averaged nuclear spin interactions yield However, an understanding of how gramicidin conducts ions information about bond angles relative to the helical axis when requires knowledge of the three-dimensional structure of fast axial rotation and coincidenceof the motional and helical gramicidin in membranes. Consequently, many techniques axes are assumed. Measurementsof gramicidin based on the have been used to obtain structural information about 13Cor lSNchemical shift interaction (Smith & Comell, 1986; gramicidin in a lipid environment. Some of the most Killian et al., 1988; Nicholson et al., 1991), however, require informative techniques have been provided by the field of independent knowledge of the orientation of the principal axis solid-state NMR.' system of the chemical shift tensor relative to the molecular In solid-state NMR studies of gramicidin in oriented frame. Furthermore, when the static shift tensor is axially bilayers, nuclear spin interactions that have been examined asymmetric, two unknown parameters must be evaluated even include the I3Cchemical shift anisotropy (Cornell et al., 1988; though measurement of a motionally averaged, chemical shift Smith et. al., 1989), the I3C-l3C dipolar interaction (Cornell powder pattern yields only a single piece of information (Seelig, et al., 1988), the lSNchemical shift anisotropy (Nicholson et 1978). In this case, such a measurement can only indicate al., 1987; Fields et al., 1988; Nicholson & Cross, 1989), the that the actual bond angle lies within a certain angular range. lsN-lH dipolar interaction (LoGrasso et al., 1989), the lSNMore definitive determinations of gramicidin bond angles I3C dipolar interaction (Teng et al., 1991), and the ZH in multilamellar dispersions are provided by measurements quadrupolar interaction (Hing et al., 1990a,b). The results basedon the 2Hquadrupolarinteraction (Datema et al., 1986; Prosser et al., 1991). Bond angle determinations are more *This work was supported by NSF Grant DIR-8720089. A.W.H. definitive in this case because the unique axis of the principal gratefully acknowledges support from the Monsanto Co. axis system of the electric field gradient (EFG) tensor is To whom correspondence should be addressed. Abbreviations: NMR, nuclear magnetic resonance; 2D REDOR, generally the same as the bond vector connected to the ZH two-dimensionalrotational-echodouble resonance;CP, cross polarization; atom and because the EFG tensor is effectively axially MAS,magic-angle spinning; CPMAS, cross-polarization magic-angle symmetric thereby requiring that only a single unknown spinning; T,,rotorperiod,DMPC,dimyristoylphosphatidylcholiie;HPLC, parameter be evaluated from a single measurement (Seelig, high-performance liquid chromatography; TLC, thin-layer chromatography; TMS, tetramethylsilane. 1977). Measurements of powder patterns representing mo0006-2960/93/0432-7593$04.00/0
0 1993 American Chemical Society
7594 Biochemistry, Vol. 32, No. 29, 1993 tionally averaged quadrupolar splittings arising from exchange-labeled gramicidin molecules indicate that N-2H bonds form angles of 16O, 25O, and 35O or 90° relative to the motional axis (Datema et al., 1986). Similar measurements performed on selectively deuterated gramicidin molecules indicate that C,-2H bond angles relative to the motional axis do not deviate by more than 5 O from the values predicted by a best-fit model of a 86.3 helical dimer (Prosser et al., 1991). These results provide strong evidence that gramicidin forms a single-stranded, 86.3helical dimer in multilamellar dispersions. Furthermore, Prosser et al. have used the average of the ratio of the quadrupolar splittings of the L residues to those of the D residues to argue that the j36.3helices are righthanded. In this paper, a recently developed solid-state NMR technique is used to investigate further the structure of gramicidin in multilamellar dispersions. Specifically, the mirror-symmetric and mirror-asymmetric versions of the twodimensional rotational-echo double-resonance (2D REDOR) experiment (Gullion et al., 1988; Gullion & Schaefer, 1989) are used to measure the W-15N dipolar coupling constant of a directly bonded pair of isotopically enriched nuclei in gramicidin incorporated into multilamellar dispersions. These 2D REDOR experiments are magic-angle spinning (MAS) experiments that are designed to measure weak heteronuclear dipolar coupling constants quantitatively. The use of the dipolar interaction to study structure and dynamics has the advantage that conclusions about the dipolar tensor can be related directly to the molecular frame. These comparisons are possible because the unique axis of the dipolar tensor and the internuclear axis coincide. Furthermore, because the dipolar tensor is exactly axially symmetric, measurement of the dipolar coupling constant allows specific values instead of angular ranges to be derived for the actual bond angle. The 13C-15Nbond of gramicidin studied in this paper is the I3C-lSN peptide linkage between the Gly2 residue and the Ala3 residue in isotopically enriched Vall-gramicidin A. Gramicidin A molecules containing isotopically enriched Glyz 13C-lcarbon atoms and Ala3 15N-amidenitrogen atoms were incorporated into multilamellar dispersions composed of DMPC and studied at a temperature above the phase transition temperature of the lipid. Under these conditions, the GlyzAla3 13C-15N peptide bond undergoes motional averaging that reduces the effective strength of the l3C-15N dipolar interaction. Mirror-symmetric 2D REDOR experiments measure the magnitude of this effective 13C-15N dipolar coupling constant in a multilamellar dispersion and the magnitude of the 13C-15N dipolar coupling constant in a powder to yield a set of two angles and their respective supplements as possibilities for the actual angle between the W-l5N bond and the helical axis. Furthermore, mirror-asymmetric 2D REDOR experiments described in this paper demonstrate how the sign of the effective 13C-15N dipolar coupling constant in a multilamellar dispersion and the sign of the 13C-15N dipolar coupling constant in a powder can be determined relative to the sign of the isotropic indirect spin-spin coupling constant. This yields information about the relative sign of these dipolar coupling constants and permits the set of possible bond angles to be reduced to a single angle and its supplement. The resulting information about the conformation of the Gly2Ala3 13C-15N peptide bond allows conclusions to be reached about the structureof gramicidin in multilamellar dispersions.
EXPERIMENTAL PROCEDURES Peptide Synthesis. Isotopic enrichment of the nuclei involved in the 13C-15N peptide linkage between the Gly2
Hing and Schaefer residue and the Ala3 residue in Vall-gramicidin A was achieved by replacing Gly2 with [l-W]Glyz and by replacing Ala3 with [15N]Ala3. The procedure used to incorporate [1J3C]glycine (99 atom '3% 13C,MSD Isotopes) and [15N]alanine(99 atom '3% 15N, MSD Isotopes) into gramicidin A and the procedure used to purify the crude peptide have been described previously (Hing et al., 1990a). The synthesis of pure Vall[ 1-13C]Glyz-[15N]Ala3-gramicidin A was verified by HPLC, TLC, 13CNMR, mass spectrometry, and amino acid analysis performed in the same fashion as described before (Hing et al., 1990a,b). Sample Preparation. The procedure for incorporatingVall[ l-13C]Glyz-[15N]Ala3-gramicidin A into multilamellar dispersions composed of DMPC followed that described for incorporating gramicidin A into oriented DMPC bilayers up to the point of hydration (Hing et al., 1990a). At this point, after the 1O:l molar ratio lipid/peptide deposits had been dried under vacuum for 2 days, the lipid/peptide deposits were placed into sample containers designed to permit liquids to be spun inside a MAS rotor. An excessof distilled,deionized water was then added to each sample container to produce a molar ratio of water to DMPC in the 80:l to 110:1 range. The samples were then repeatedly subjected to a cycle consisting of vortexing, freezing, and thawing. Finally, the samples were heated at 43 OC for at least 4 days. NMR NMR Console and Probe. Observation of 13Cnuclei was performed with a home-built spectrometer that is capable of producing radio frequency pulses at three different frequencies with the field strength of the pulses regulated from scan to scan. The NMR probe used to observe the 13C signals was a triply-tuned, transmission line probe built by R. A. McKay (US.Patent 4,446,431). The probe allows high-power radio frequency pulses to be applied at 200 (lH), 50 (13C),and 20 MHz (15N). The probe also allows samples to be spun at the magic-angle at speeds up to 4 kHz. Observation of 31P nuclei was performed with a similar, four-channel instrument that allows signals at 81 MHz (31P) to be observed. Temperature control was achieved by regulating the temperature of the air that drives the journal bearings. For samples in powder form, data were acquired at room temperature. For multilamellar dispersions, data were acquired at a temperature well above the phase transition temperature of approximately 28 OC (Chapman et al., 1974, 1977; Nicholson et al., 1987). 31PNMR. Phosphorus NMR data of a nonspinningsample of multilamellar dispersions were obtained with a phase-cycled Hahn-echo experiment (Rance & Byrd, 1983). The 31P 90° pulse width was 6.6 ps, the interpulse delay was 24 ps, therecycle delay was 1 s, and the strength of the lH decoupling field was 60 kHz. Spectra were acquired with a spectral width of 10 000 Hz and a spectral size of 2048 complex points and were plotted with a line broadening of 80 Hz. The 0 ppm reference point of the spectra was set equal to the isotropic resonance frequency, as determined by room temperature 31P CPMAS experiments, of external DMPC in powder form. I3C CPMAS. Carbon- 13 CPMAS spectra were acquired at a MAS speed of l/Tr = 1000 Hz with quadrature phase cycling and spin temperature alternation. Spectra were referenced to an external TMS standard and were acquired with a spectral width of 20 000 Hz and a recycle delay of 1 s. For samples in powder form, cross polarization from lH to 13C was performed at a field strength of 35 kHz with a contact time of 2 ms; the strength of the 'H decoupling field
Biochemistry, Vol. 32, No. 29, 1993 7595
#
C
ISN
.
d
,IsN
Y
*
"
"
.
-
"
"
flqq311 1 1 qqqqq9qq l
.
Y
I 1 I I 0 I 2 3
V
'
l
I
I
4
5
I
.
I
.
V
.
I
Y
'
"
' *
.
V
.
Y
Y
.
"
X
I
I
I
qq7111111111 J
I
I
I
I
I
I
I
I
I
I
6 7 8 9 IO 11 12 13 14 15 16 17 18 19 20 T,
FIGURE1: 2D REDOR pulse sequences. The mirror-symmetric experiment with (a) 8 and (c) 16 rotor periods of effective dephasing and the mirror-asymmetric experiment with (b) 8 and (d) 16 rotor periods of effective dephasing are diagrammed.
was 100 kHz; the spectral size was 1024 complex points; and line broadening of 40 Hz was used to process spectra. For samples of multilamellar dispersions, cross polarization from IH to 13Cwas performed at a field strength of 38 kHz with a contact time of 2 ms; the strength of the 'H decoupling field was 60 kHz; the spectral size was 2048 complex points; and line broadening of 20 Hz was used to process spectra. 13C-Observed,I5N-Dephased 2D REDOR. The mirrorsymmetric and mirror-asymmetric 2D REDOR pulse sequences used in this paper are diagrammed in Figure 1. The initial cross polarization from protons to carbons is followed by a train of 13Cu pulses each of which is applied at the end of a rotor period, a train of 15N a pulses each of which is applied at a time tl from the start or end of a rotor period, and high-power proton decoupling. In the mirror-symmetric sequences (Figure la,c), the second set of 15N a pulses is incremented in a direction opposite to that of the first set. In the mirror-asymmetric sequences (Figure 1b,d), all the 15N 'K pulses are incremented in the same direction. The application of a train of u pulses to both I3C and 15N nuclei and phase alternation of each a pulse train according to an XY-8, compensated Carr-Purcell scheme (Gullion et al., 1990) are modifications of the original 2D REDOR sequence (Gullion et al., 1988; Gullion & Schaefer, 1989) that have been described before (Gullion et al., 1990). The addition of a 13Ca pulse (denoted by II in the sequences of Figure 1) to the XY-8train of I3C ?r pulses is required for full implementation of the XY-8phase alternation scheme. Over the time period beginning one rotor period before the II pulse and ending one rotor period after the II pulse, the net phase accumulated due to isotropic and anisotropic chemical shift interactions and heteronuclear dipolar and indirect spin-spin interactions is zero. Therefore, the total number of rotor periods of dephasing is two less than the total number of rotor periods between the 13C CP pulse and data acquisition. Acquisition of the 13Cfree induction decay occurs during the second time dimension t2. Two-dimensional REDOR spectra were acquired at a MAS speed of l/Tr = 1000 Hz with the 13C and 15N carrier frequencies placed on resonance for the nuclei of interest. The phases of the XY-8a pulses were left unchanged while phase cycling of the 13C CP pulse, the ll pulse, and the receiver
followed the scheme of Rance & Byrd, 1983. The spin temperature was alternated between consecutive executions of this phase-cycling scheme. Cross polarization from 'Hto I3Cwas performed at a field strength of 38 kHz with a contact time of 2 ms. The 90° pulse width was 6.6 ps for 'H, 13C, and 15N, The 13C spectra were referenced to an external TMS standard and were acquired with a spectral width of 20 OOO Hz, a spectral size of 2048 complex points, and a recycle delay of 1 s. For samples in powder form, 2D REDOR spectra were acquired with Nc = 8 rotor periods of effective dephasing (Figure la,b); 15N'K pulse placement as measured by $1 was incremented in steps of Tr/50;the strength of the 'Hdecoupling field was 100 kHz;and line broadening of 40 Hz was used to process spectra. For samples of multilamellar dispersions, 2D REDOR spectra were acquired with Nc = 16rotor periods of effective dephasing (Figure lc,d); 15Nu pulse placement was incremented in steps of Tr/32; the strength of the 'H decoupling field was 60 kHz; and line broadening of 20 Hz was used to process spectra. For powder samples and multilamellardispersions, a greater number of scans was used to acquire 2D REDOR spectra when magnetization was observed to change rapidly as a function of tl because magnetization behavior in this t l time-domain region contains most of the information about the value of the dipolar coupling constant.
THEORY Basic Principles. The 2D REDOR experiment is based upon the creation of an average dipolar interaction and an average indirect spin-spin interaction between two heteronuclear spins, I and S, under conditions of MAS (Gullion & Schaefer, 1989). These average interactions are created by the appropriate application of a pulses. The spin system then evolves under the influence of the average dipolar and indirect spin-spin interactions at an average angular frequency that is determined by the placement of the ?r pulses. This average angular frequency of evolution ultimately determines the amount of magnetization that is observed. Powder Samples. In the relatively immobile environment of a powder, the observable magnetization Ms for an isotropic distribution of I S internuclear vector orientations is obtained by summation of the magnetization components for a single orientation and is given by
where &(a,B) is the average angular frequency of evolution for the a,B orientation in rad/s, Nc is the total number of rotor periods of dephasing, and TIis the rotor period in seconds. The symbols a and B are the azimuthal and polar angles, respectively, of the I S internuclear vector's initial orientation relative to the rotor axis. Since &(a,@)is determined by the placement of the u pulses, the average angular frequency created by the mirrorsymmetric 2D REDOR sequence (Gullion et al., 1988;Gullion & Schaefer, 1989) &(a,@)= f((D/2)[2.\/2 sin 2P sin a [cos artl - 11 -
sin2B sin 2a [cos 2wrt1- 111) (2) is different from the average angular frequency created by the mirror-asymmetricsequence (original equation given by Gullion & Schaefer, 1989; presented here in corrected form)
7596 Biochemistry, Vol. 32, No. 29, 1993
a(a,fi)= f { ( D / 2 ) [ 2 & sin 28 [sin(a + tortl) - sin a ] sin’ fi [sin(2a + 2w,t1) -sin 2a11 + J(w,t, - a)) (3) where D is the coupling constant that characterizes the dipolar interaction between the I spin and S spin in the powder (in hertz), and J is the coupling constant that characterizes the isotropic indirect spin-spin interaction between those same two spins (in hertz). The symbol o r = 2 a / Tr is the MAS speed in rad/s, and tl measures the position of the lSNa pulses in seconds. Examination of eqs 1-3 shows that the average angular frequency’s magnitude lij(a,/3)1 and therefore the observable magnetization MSreflect only the magnitude of D when the mirror-symmetric sequence is applied to a powder sample but reflect both the magnitudes of D and J and their relative sign when the mirror-asymmetric sequence is applied. In principle,the puredipolar coupling constant for absolutely motionless molecules can be calculated from the following equation: D = yIysh/4a’r3 (4) where D is in hertz, 71is the gyromagnetic ratio of the I spin, ys is the gyromagnetic ratio of the S spin, h is Planck’sconstant, and r is the internuclear distance. However, measured values of D often differ from theoretical values because of the presence of motion or other factors. Multilamellar Dispersions. In the liquid-crystallinephase of multilamellar dispersions,the I S internuclearvector rotates rapidly about a motional axis. The observable magnetization for such a system can be calculated by unitary transformation of the dipolar tensor from its principal axis system to the motional axis frame, transformation from the motional axis frame to the rotor frame, and then transformation from the rotor frame to the laboratory frame. The resulting equations can be simplified because the I S internuclear vector rotates about the motional axis at a rate (>>lo5 Hz) much greater than typical MAS speeds (- lo3Hz) and much greater than typical values of weak dipolar coupling constants (S103 Hz). In this case, the observable magnetization MSfor an isotropic distribution of motional axis orientations is given by
where &(&a’,@’)is the average angular frequency of evolution for the B,a’,j3’ orientation in rad/s. The symbol B is the polar angle between the I S internuclear vector and the motional axis, and the symbols a’ and are the azimuthal and polar angles, respectively, of the initial orientation of the motional axis relative to the rotor axis. The average angular frequency a(O,a’,F) created by the mirror-symmetric 2D REDOR sequence in a system undergoing fast rotation about a motional axis is given by
a(B,a’,p’) = f { ( D e / 2 ) [ 2 & sin 2Fsin a’ [cos artl- 11 sin’ F s i n 2a’ [cos 2wrt1- 111) ( 6 ) while the average angular frequency G(O,a’,F) created by the mirror-asymmetric sequence is given by
a(B,a,’F) = f { ( D e / 2 ) [ 2 & sin 28’ [sin(a’ + w,tl) sin a’] - sin’ 8’ [sin(2a’ + 20,t1) - sin w]] + ~ ( w , t , ?r)) (7) The effective coupling constant De (in hertz) depends on B and characterizes the effective strength of the dipolar interaction between the I spin and S spin in systems undergoing
Hing and Schaefer fast rotation about a motional axis. As before, the coupling constant J characterizes the isotropic indirect spin-spin interaction between those same two spins. Examination of eqs 5-7 shows that the average angular frequency’s magnitude Ia(B,a’,fl’)l and therefore the observable magnetization MS are determined by the magnitude of De when the mirrorsymmetric sequence is applied and are determined by the magnitudes of Deand Jand their relative sign when the mirrorasymmetric sequence is applied. The equation for De in terms of B is given by De = D(3 cos’ 8 - 1)/2
(8) which permits the effective dipolar coupling constant in the membrane, De, to be related directly to the dipolar coupling constant in the powder, D. Clearly, De and D can possess identical or opposite signs depending on the value of 8. The use of the powder dipolar coupling constant D in eq 8 as the reference for calculating De is predicated on two assumptions. The first assumption is that the I S internuclear distance is the same in the liquid-crystallinephase and the powder, This assumption is certainly true for a one-bond I S spin pair. The second assumption is that small-amplitude, high-frequency motions in the liquid-crystalline phase do not differ significantly from those in the powder and therefore do not cause the reference dipolar coupling constant to differ significantly from D. This is certainly a valid premise if eq 8 is used to calculate values for the angle B when De
